Part-Level 3D Gaussian Vehicle Generation with Joint and Hinge Axis Estimation
Summary: arXiv:2604.05070v1 Announce Type: new
Abstract
Simulation is essential for autonomous driving, yet current frameworks often model vehicles as rigid assets and fail to capture part-level articulation. With perception algorithms increasingly leveraging dynamics such as wheel steering or door opening, realistic simulation requires animatable vehicle representations. Existing CAD-based pipelines are limited by library coverage and fixed templates, preventing faithful reconstruction of in-the-wild instances.
Innovative Approach
We propose a generative framework that synthesizes an animatable 3D Gaussian vehicle from a single image or sparse multi-view input. Our method addresses two critical challenges:
- Challenge 1: Large 3D asset generators are typically optimized for static quality, failing to consider articulation. This leads to distortions at part boundaries when animated.
- Challenge 2: Segmentation alone cannot provide the kinematic parameters required for motion, which is essential for realistic simulation.
Proposed Solution
To overcome these challenges, we introduce two key components:
- Part-Edge Refinement Module: This module enforces exclusive Gaussian ownership at the edges of vehicle parts, ensuring that transitions between parts are smooth and visually coherent.
- Kinematic Reasoning Head: This component predicts joint positions and hinge axes of movable parts, which are crucial for animating the vehicle accurately.
Benefits of the Framework
Together, these components enable a faithful part-aware simulation, bridging the gap between static generation and animatable vehicle models. The implications of this research extend beyond mere aesthetics; they are poised to enhance the capabilities of autonomous driving systems significantly.
Conclusion
This approach marks a significant advancement in vehicle simulation technology, addressing the limitations of existing frameworks and paving the way for more sophisticated and realistic autonomous driving simulations. By providing animatable representations of vehicles, our framework supports the development of advanced perception algorithms that can react to dynamic environments.
The future of autonomous driving simulations looks promising as we continue to refine these methods and explore new avenues for improvement.